設備異常診断用の時系列データ検索言語TPQL

抄録

In facility management for plants and buildings, needs of facility diagnosis for saving energy or facility management cost by analyzing time series data from sensors of equipments in facilities have been increasing. In this paper, we propose a relation-based query language TPQL (Trend Pattern Query Language) for expressing constraints in time series data for anomaly detection in facilities and implemented an anomaly detection system based on TPQL. The features of TPQL are the following. (1) TPQL introduces a convolution operator into SQL (Structured Query Language) in order to describe contextual anomaly conditions over window sequences such as duration constraint and hunting constraint. (2) TPQL introduces time-interval based join into SQL in order to join time series data with different sampling rates. The anomaly detection system consists of a TPQL-interpreter as a top-level engine, relational database as an SQL engine, a key-value store database as a large data storage and configure management information to represent target signals for diagnosis and threshold values for anomaly detection. We evaluate that the system has enough expression ability to describe domain dependent anomaly detection conditions with TPQL over sliding windows and the sufficient processing speed required by the real applications.